Dexova AIDexova AI

Five AI features that lift your online store’s sales.

Dexova AI adds five AI-powered sales features to your online store with one snippet: product recommendations, semantic search, automatic upsell, a product chatbot, and product comparison. All of them run on top of your own catalog — no replatforming, no integration project, and we never touch your payments.

Product recommendations

AI product recommendations that surface automatically

Vector similarity over your own catalog puts "related" and "for you" rows on every product page — no rules to maintain, no LLM cost per request.

For youWidget mockup
Arka Trail Shoes
Rp 489.000
AI recommended
Dara Merino Tee
Rp 259.000
How it works
Each product is turned into an embedding once, at index time. When a shopper opens a product, the engine finds its nearest neighbours in vector space — the products closest in meaning, not just the same category.
The impact
Shoppers see more relevant products every visit, so more of them add to cart. Because it runs on vector search — not an LLM per click — cost stays flat as traffic grows.
Where it appears
Home, product, and cart pages — as "for you" and "related products" rows.
Automatic upsell

Automatic upsell that grows the cart

At the cart, the engine pairs vector similarity with your margin and bundle rules — the suggestion that lifts order value, right before checkout.

Cart · 1 itemWidget mockup
Arka Trail ShoesRp 489.000
Often bought together
Trail Socks — 3 pairsRp 89.000+ Add
How it works
When a shopper adds an item, the engine finds complementary products often bought together, then filters them by the margin and stock rules you set.
The impact
Average order value rises without aggressive discounting. The suggestion appears exactly when the buying decision is hot — not in a follow-up email that gets ignored.
Where it appears
The cart page and checkout step, as an "often bought together" block.
Product chatbot

An online-store product chatbot that answers 24/7

RAG over your product info and policies, with an anti-hallucination guardrail: it answers from your catalog or says it doesn’t know — never invents.

Product assistantWidget mockup
Is this fabric cool enough for daily wear?
Yes — the Dara Merino Tee uses 17.5-micron merino wool that wicks sweat and stays cool for everyday wear. Great for tropical climates.
Source: product descriptionAnswers only from your catalog
How it works
The shopper’s question is matched to the most relevant chunks of your product info and policies, then the model summarises them into an answer. Price and stock always come from live data — never made up.
The impact
Pre-sale questions get answered 24/7 without you on standby. Accurate answers reduce hesitation — and reduce returns from mismatched expectations.
Where it appears
A floating chat widget on product and cart pages, rendered in a Shadow DOM so it won’t break your theme.
Product comparison

Product comparison that helps shoppers decide

Two or three products, compared by the attributes that matter — a clear verdict instead of an abandoned tab.

Compare 2 productsWidget mockup
Trail Arka
Halim Feather
Weight
240 g
198 g
Terrain
Trail
Road
Price
Rp 489.000
Rp 429.000
AI: For trail running, Arka is the better pick — its grip and protection are built for that terrain. Halim wins if weight on the road is your priority.
How it works
The engine pulls structured attributes from your product data, lays them out in a side-by-side table, then adds a one-sentence verdict: which product for which need.
The impact
Shoppers torn between two products decide faster, on your page — instead of opening another tab to research and forgetting to come back.
Where it appears
Product pages and search results, when a shopper flags two or more products.

Behind the one snippet.

Five features, one engine. Here’s what keeps it fast, cheap to run, and safe for your catalog.

Vector search, not an LLM per click

Recommendations, search, and upsell run on vector search — not an LLM call per request. That keeps latency low and cost flat as traffic grows.

Embed once, sync smart

Your catalog is embedded once. When a product changes, only the changed product is re-embedded — not the whole catalog. Catalog updates stay cheap and fast.

A chatbot that can’t make things up

The chatbot only answers from your catalog and policies. If the answer isn’t there, it says it doesn’t know — price and stock are never invented.

Every store isolated at the database level

Each merchant’s catalog is separated with row-level security. Your data is never visible to another store, and never used to answer for one.

FAQ

Feature questions

Do I need to change my theme or platform?

No. Every feature installs with one snippet and renders inside a Shadow DOM — your theme is untouched, and features run on any stack that can load a script tag.

Which platforms are supported?

Any stack that renders HTML: custom builds, WooCommerce, Shopify, or a static site. If your store can load one script tag, the features work.

Can the chatbot get things wrong or make answers up?

No. The chatbot only answers from your catalog and policies, with an anti-hallucination guardrail. If the answer isn’t there, it says it doesn’t know — price and stock are never invented.

How long until the features are live?

About 15 minutes: sign up, upload your catalog via CSV or API, then paste the snippet. The engine indexes your catalog and the features start working.

Start selling more this week.

Free to start. One snippet. Your dashboard shows the lift.

Get startedHave us build it

You keep your payments — we sell software, we never touch your transactions.